05.26: Reliability assessment for local buckling of plates

Abstract

ABSTRACTIn the current Eurocode 3‐1‐5 [10] for local buckling, the resistance curve used to represent the reduction factor of plated elements due to local failure is based on the so‐called “Winter‐curve”, which was derived using a semi‐empirical approach by George Winter in 1947. This design curve reproduces the mean reduction values achieved in the experiments conducted by Winter and other researchers. However, when applying the safety concept of the Eurocode (EN 1990) [8], an additional safety factor is necessary to ascertain a defined level of failure probability. Currently, this factor is set to 1.0 for applications in building structures.In the study at hand, 131 stub column tests on welded, squared box sections from steel grades S275 up to S960 were evaluated to assess a realistic safety factor in conformity with the safety standard EC0 [8]. A considerable amount of tests where thereby classified as high strength steel. The results showed to be comparable for all steel grades in dependence of their local slenderness.The evaluation of test results revealed a considerable scatter, which is partly attributed to the stability failure mode, but could be also found to be originated from lack of information concerning unintended eccentricities in the experimental test setups. Therefore, the data set was divided into two sets, one consisting of all data, and one including only tests where these information were available and thus had reduced scatter. The second set lead consequently to a more favourable safety factor than the first one.For all sets using the procedure to derive a safety factor, assumptions have to be made regarding the property‐specific Coefficients of Variation (CoVs). Standard values often given in literature, although correct for their specific applications, cannot be simply adjusted when including high strength steel material. Additionally, as the resistance function for local buckling is complex, the impact of e.g. the thickness has to be taken into account carefully. Different approaches and variations are presented in this paper, leading to scientifically justified ‐values.